Even if you're looking for causation, detrending is usually necessary to obtain consistent estimators (in statistical terms). For example, Granger causality wor

Even if you're looking for causation, detrending is usually necessary to obtain consistent estimators (in statistical terms). For example, Granger causality works with two stationary timeseries: https://en.wikipedia.org/wiki/Granger_causality https://en.wikipedia.org/wiki/Granger_causality